Groupe des Utilisateurs Francophones de CDISC Bagneux - June 20th, 2008 Laurence Le Lann,

1. Overview and feed-back from CDISC European Interchange 2008(From April 21st to 25th, COPENHAGEN) Groupe des Utilisateurs Francophones de CDISCBagneux - June 20th, 2008 Laurence Le Lann, Hélène Savoye

2. Agenda • SDTM Training • ADaM Training • Messages from SDTM & ADaM training • Legacy Data Conversion Workshop • General comments from that Interchange • Useful WEB links • Conclusion

3. SDTM Training - Content of 2 full days(1/2) • Review the mains terms used by CDISC, their definition and the differences between some of them (SDTM vs AdaM…) • Annotate the blank CRF with SDTM variables • How to choose the correct model • How to create a new domain (finding, intervention or event) • Create the defines for SDTM • list of variables • name of variables …. • no use of derived variables (they should be kept in ADS)

4. SDTM Training - Content of 2 full days(2/2) • Format for Date/Time variables • Create few SUPPQUAL • Create some Trial design define (Subject level data, Subject Elements and Subject visits) • Create tables for “relationships”

5. ADaM Training - Content of 1 full day • Background and introduction of CDISC and Standards, CDISC and the FDA • Purpose of ADaM • Basic principles of ADaM • ADaM rules and variables • ADaM Metadata • ADSL • ADaM Implementation guide - Draft

6. Messages from SDTM & ADaM training (1/7) • FDA according to CDISC presenter’s experience • FDA wants to be able to replicate what the sponsor did so that they can trust them • First thing FDA statistician do • Recreate analyses datasets • Some tools at FDA only use SDTM • Some derived variables are needed in SDTM as well • ie “total score for a questionnaire” • FDA does not employ programmers • Statistical and Medical reviewers need ADaM • Traceability is of high importance • Traceability from p-value back to protocol • CDISC recommends to QC the transparency so that FDA understand easily what the sponsor did and how he did it • Keep in touch with FDA reviewer • to understand what they want • So that they understand what you did

7. Messages from SDTM & ADaM training (2/7) • Up to sponsor, but CDISC Recommendation • Use SDTM to build ADaM • Separate submission of SDTM and ADaM • With corresponding define files • No real recommendation from CDISC on derived variables to be in SDTM • “Include some derived as they seem natural to be in SDTM”

8. Messages from SDTM & ADaM training (3/7) • Distinguish between what was observed and what was derived • No imputation should be done in SDTM • If imputation needed, should be in ADaM • Analysis Timing variables • Character variables (C) in SDTM • Numerical variables (N) in ADaM • if lots of imputation in (N) variable, keep (C) in ADaM as well CDISC suggestion to FDA: Minimum mandatory submission should include SDTM and ADSL from ADaM (still in discussion with FDA)

9. Messages from SDTM & ADaM training (4/7) • Specifications for organizing the datasets in eCTD • Contents of ECTD folders: • Folder Datasets/[study]/analysis => should contain analysis datasets and corresponding define files • Folder Datasets/[study]/tabulations => should contain data tabulation datasets, corresponding define files and blank annotated CRF • Annotated CRF recommended in data tabulation datasets folder with links in other directories (listing, analysis) • Recommendation is to not refer to annotated CRF in ADaM but to SDTM which themselves do refer to annotated CRF

10. Messages from SDTM & ADaM training (5/7) • If submission of SDTM, submission of subject profiles might not be needed • Need agreement with FDA at pre-NDA meeting • Permissible variables • if information collected in CRF, variables must be submitted in SDTM • (U?)SUBJID: • Recommendation from CDISC is • The SUBJID uniquely identify each subject within a submission and ensure that the same person has the same number across studies • However, Not always possible • Enough information might not be allowed to be collected to ensure same numbering across studies within submission • Information on randomization (IVRS) • Might be recorded in sub-qualifier DM • However never discussed in SDTM WG • Trial design domain • Not required but recommended for clinical trials

11. Messages from SDTM & ADaM training (6/7) • Pilot ADaM submission • Not always what you are supposed to do! • “Role” column in define • Was present in ADaM 2.0 • Not included anymore in 2.1 since variables may have different roles within analyses • “Reviewer guide” • Optional document presenting key decisions • Should add additional information compared to SAP • Not a CDISC document, no template exists • FDA’s feedback in ADaM pilot: Felt as very Helpful • ADaM 2.1 version just released for public consultation/review by September 5, 2008

12. Messages from SDTM & ADaM training (7/7) • “one proc away” table • Dose not mean only one SAS proc to produce a given table • A table containing mixed information (baseline, value at time X, change from baseline and lsmeans estimates) OK • as long as information to produce this table available somewhere • even if you need 4 proc to produce the table • Program codes • “Ensure program codes are well commented, easy to follow and well formed” • Provide code to help reviewer understand what you did (ie PROC MIXED with options) • FDA does not expect to run the code, just need to understand the options used

13. Legacy Data Conversion Workshop • A ½ day workshop with 4 examples about the way to convert legacy data into SDTM data • Very interesting workshop • Presentation of process to convert legacy data • Retrieve of all information needed • Harmonization of controlled terminology • CRF annotation • Mapping • QC • Demo of SAS ETL STUDIO used to do the mapping • Link to WEB site will be given later?

14. General comments from that Interchange(1/3) • Gray zones in SDTM definition on derived variables • Part of SDTM? SUPPQUAL? ADaM? • Different approaches/interpretations of SDTM IG / ADAM but All the presentations report the data flow SDTM -> ADS • Very wide profiles of participants (50% DM) • A lot of new standards are under definition at CDSIC level, for Data management mainly • Operational Data Model (ODM) trends to be a standard way to store data from EDC? • Trend to integrate all kind of data used in research into 1 standard (public health as well)

15. General comments from that Interchange(2/3) • Risks • Not follow-up the discussions and orientations • Not easy to understand all the initiatives: HL7, ODM etc …. • Key words for 2008 • TRANSPARENCY & INTEROPERABILITY • Important focus on QC aspects, SDTM adherence checks • Importance of metadata • Ensure that FDA reviewers have complete & accurate information

16. General comments from that Interchange(3/3) Training for all DM and PROG staff to SDTM is very important (could be organized internally) Training for all STAT staff to AdAM is also recommended

17. Useful WEB links • On FDA websites: • Checks performed by the FDA before any loading attempt in Data warehouse JANUS: http://www.fda.gov/oc/datacouncil/janus_sdtm_validation_specification_v1.pdf • Study data specifications: http://www.fda.gov/cder/regulatory/ersr/studydata.pdf • On CDISC website: • Presentations from Congress part: http://www.cdisc.org/publications/interchange2008.html

18. Conclusion Very interesting congress & trainings Recommended to EVERYBODY !